AI mock interview platform — adaptive follow-ups, voice & video answers, and measured delivery scoring.
- Architected an end-to-end AI interview platform — dynamically generates interviews from job descriptions, résumés, custom topics, or behavioral (cultural-fit) tracks with adaptive AI follow-ups, supporting text, voice, and video-recorded responses.
- Engineered a quantitative voice-analytics pipeline — built a dedicated FastAPI + Parselmouth microservice that measures pace, pauses, filler words, pitch, and energy directly from raw audio, delivering reports with per-question scoring, model ideal answers, and personalized improvement guidance.
- Designed fault-tolerant, secure infrastructure — implemented Gemini with automatic Groq failover via rotating API key pools (exponential backoff on 429/5xx), Groq Whisper speech-to-text, an idempotent self-healing report pipeline, and Argon2 + JWT (jose) authentication with strict per-user data isolation.






